# Computations
import numpy as np
import pandas as pd

import catboost

# Sklearn
from sklearn import metrics
from sklearn import preprocessing
from sklearn.impute import SimpleImputer

# Visualisation libraries

## Text
from colorama import Fore, Back, Style
from IPython.display import Image, display, Markdown, Latex, clear_output

## progressbar
import progressbar

## plotly
from plotly.offline import init_notebook_mode, iplot 
import plotly.graph_objs as go
import plotly.offline as py
from plotly.subplots import make_subplots
import plotly.express as px

## seaborn
import seaborn as sns

## matplotlib
import matplotlib.pyplot as plt
from matplotlib.patches import Ellipse, Polygon
from matplotlib.font_manager import FontProperties
import matplotlib.colors as mcolors
from matplotlib import cm
plt.style.use('seaborn-whitegrid')
plt.rcParams['axes.labelsize'] = 14
plt.rcParams['xtick.labelsize'] = 12
plt.rcParams['ytick.labelsize'] = 12
plt.rcParams['text.color'] = 'k'
%matplotlib inline

import warnings
warnings.filterwarnings("ignore")

Home / Healthcare Analytics and Modeling

Liver Disorders Dataset

In this article, we analyze the Liver Disorders Dataset from the UCI Machine Learning Repository.

Picture Source: niddk.nih.gov

Dataset Information:

The first five variables are all blood tests thought to be sensitive to liver disorders that might arise from excessive alcohol consumption. Each line in the dataset constitutes the record of a single male individual.

Important note: The 7th field (selector) has been widely misinterpreted in the past as a dependent variable representing the presence or absence of a liver disorder. This is incorrect [1]. The 7th field was created by BUPA researchers as a train/test selector. It is not suitable as a dependent variable for classification. The dataset does not contain any variable representing the presence or absence of a liver disorder. Researchers who wish to use this dataset as a classification benchmark should follow the method used in experiments by the donor (Forsyth & Rada, 1986, Machine learning: applications in expert systems and information retrieval) and others (e.g. Turney, 1995, Cost-sensitive classification: Empirical evaluation of a hybrid genetic decision tree induction algorithm), who used the 6th field (drinks), after dichotomizing, as a dependent variable for classification. Because of widespread misinterpretation in the past, researchers should take care to state their method clearly.

Attribute Information:

Attribute	Information
MCV	Mean corpuscular volume
AlkPhos	Alkaline Phosphotase
Sgpt	Alamine Aminotransferase
Sgot	Aspartate Aminotransferase
GammaGT	Gamma-Glutamyl Transpeptidase
Drinks	Number of half-pint equivalents of alcoholic beverages drunk per day
Selector	Field used to split data into two sets

Data = np.genfromtxt('liver-disorders/bupa.data', delimiter=',', dtype = int)
Attributes = ['MCV', 'AlkPhos', 'Sgpt', 'Sgot', 'GammaGT', 'Drinks', 'Selector']
Data = pd.DataFrame(data = Data, columns = Attributes)
Target = 'Drinks'
display(Data.head())

def Data_Plot(Inp, Title = None, W = None):
    data_info = Inp.dtypes.astype(str).to_frame(name='Data Type')
    Temp = Inp.isnull().sum().to_frame(name = 'Number of NaN Values')
    data_info = data_info.join(Temp, how='outer')
    data_info ['Size'] = Inp.shape[0]
    data_info['Percentage'] = 100 - np.round(100*(data_info['Number of NaN Values']/Inp.shape[0]),2)
    data_info = data_info.reset_index(drop = False).rename(columns = {'index':'Features'})
    #
    fig = px.bar(data_info, x= 'Features', y= 'Percentage', color = 'Data Type',
                 text = 'Percentage',
                 color_discrete_sequence = ['PaleGreen', 'LightCyan', 'PeachPuff', 'Pink', 'Plum'],
                 hover_data = data_info.columns)
    fig.update_layout(plot_bgcolor= 'white', legend=dict(x=1.01, y=.5, traceorder="normal",
                                                         bordercolor="DarkGray", borderwidth=1))
    if not W == None:
        fig.update_layout(width = W)
    fig.update_traces(texttemplate= 10*' ' + '%%{text}', textposition='inside')
    fig.update_traces(marker_line_color= 'Black', marker_line_width=1., opacity=1)
    if not Title == None:
        fig.update_layout(title={'text': '<b>' + Title + '<b>', 'x':0.5,
                                         'y':0.90, 'xanchor': 'center', 'yanchor': 'top'})
    fig.show()
    return data_info

data_info = Data_Plot(Data, Title = 'Liver Disorders Dataset', W = 500)

	MCV	AlkPhos	Sgpt	Sgot	GammaGT	Drinks	Selector
0	85	92	45	27	31	0	1
1	85	64	59	32	23	0	2
2	86	54	33	16	54	0	2
3	91	78	34	24	36	0	2
4	87	70	12	28	10	0	2

Problem Description

In this article, the dependent variable is the number of drinks. Note that, Selector column is intended to split the data into train and test subsets for one particular experiment.

Features with high variance

Moreover, high variance for some features can hurt our modeling process. For this reason, we can standardize features by removing the mean and scaling to unit variance.

def Feature_Normalize(X, PD):
    def List_Break(mylist, n = PD['word_break']):
        Out = []
        for x in mylist:
            y = x.split()
            if len(y)> n:
                z = ' '.join(y[:n])
                sep = np.arange(0, len(y), n)[1:]
                for n in sep:
                    z = z + '\n'+ ' '.join(y[n:])
            else:
                z = ' '.join(y)
            Out.append(z)
        return Out

    X = Data.drop(columns = [Target])
    y = Data[Target]

    scaler = preprocessing.StandardScaler()
    X_std = scaler.fit_transform(X)
    X_std = pd.DataFrame(data = X_std, columns = X.columns)

    fig, ax = plt.subplots(2, 1, figsize = PD['figsize'])
    ax = ax.ravel()
    CP = [sns.color_palette("OrRd", 20), sns.color_palette("Greens", X.shape[1])]
    Names = ['Variance of the Features', 'Variance of the Features (Standardized)']
    Sets = [X, X_std]
    kws = dict(label='Feature\nVariance', aspect=10, shrink= .3)
    for i in range(len(ax)):
        Temp = Sets[i].var().sort_values(ascending = False).to_frame(name= 'Variance').round(2).T
        _ = sns.heatmap(Temp, ax=ax[i], annot=True, square=True,  cmap = CP[i],
                        linewidths = 0.8, vmin=0, vmax=Temp.max(axis =1)[0],  annot_kws={"size": PD['annot_text_size']},
                        cbar_kws=kws)
        if not PD['word_break'] == None:
            mylist = List_Break(Temp.T.index.tolist())
            _ = ax[i].xaxis.set_ticklabels(mylist)
        _ = ax[i].set_yticklabels('')
        _ = ax[i].set_title(Names[i], weight='bold', fontsize = 14)
        _ = ax[i].set_aspect(1)
        del Temp
    plt.subplots_adjust(hspace=PD['hspace'])
    Out = pd.DataFrame(data = X_std, columns = X.columns.tolist())
    return Out

X = Data.drop(columns = [Target])
y = Data[Target]

PD = dict(figsize = (8, 7), hspace = 0.2, annot_text_size = 12, word_break = 2)
X = Feature_Normalize(X, PD)

Train and Test sets

def Header(Text, L = 100, C = 'Blue', T = 'White'):
    BACK = {'Black': Back.BLACK, 'Red':Back.RED, 'Green':Back.GREEN, 'Yellow': Back.YELLOW, 'Blue': Back.BLUE,
         'Magenta':Back.MAGENTA, 'Cyan': Back.CYAN}
    FORE = {'Black': Fore.BLACK, 'Red':Fore.RED, 'Green':Fore.GREEN, 'Yellow':Fore.YELLOW, 'Blue':Fore.BLUE,
         'Magenta':Fore.MAGENTA, 'Cyan':Fore.CYAN, 'White': Fore.WHITE}
    print(BACK[C] + FORE[T] + Style.NORMAL + Text + Style.RESET_ALL + ' ' + FORE[C] +
          Style.NORMAL +  (L- len(Text) - 1)*'=' + Style.RESET_ALL)
    
def Line(L=100, C = 'Blue'):
    FORE = {'Black': Fore.BLACK, 'Red':Fore.RED, 'Green':Fore.GREEN, 'Yellow':Fore.YELLOW, 'Blue':Fore.BLUE,
         'Magenta':Fore.MAGENTA, 'Cyan':Fore.CYAN, 'White': Fore.WHITE}
    print(FORE[C] + Style.NORMAL + L*'=' + Style.RESET_ALL)

def Search_List(Key, List): return [s for s in List if Key in s]

def DataSize(Inp):
    return pd.DataFrame({'Number of Instances': [Inp.shape[0]], 'Number of Attributes': [Inp.shape[1]]}).style.hide_index()

# Train Set
Header(Text = 'Train')
Train = Data.loc[Data.Selector == 2].drop(columns = 'Selector').reset_index(drop = True)
display(DataSize(Train))
X_train = Train.copy()
y_train = X_train.pop(Target)

# Test Set
Header(Text = 'Test', C = 'Green')
Test = Data.loc[Data.Selector == 1].drop(columns = 'Selector').reset_index(drop = True)
display(DataSize(Test))
X_test = Test.copy()
y_test = X_test.pop(Target)
Line()

display(pd.DataFrame(data={'Set':['X_train','X_test','y_train','y_test'],
               'Shape':[X_train.shape, X_test.shape, y_train.shape, y_test.shape]}).set_index('Set').T)

Train ==============================================================================================

Number of Instances	Number of Attributes
200	6

Test ===============================================================================================

Number of Instances	Number of Attributes
145	6

====================================================================================================

Set	X_train	X_test	y_train	y_test
Shape	(200, 5)	(145, 5)	(200,)	(145,)

Modeling: CatBoost Regressor

CatBoost AI is based on gradient boosted decision trees. During training, a set of decision trees is built consecutively. Each successive tree is built with reduced loss compared to the previous trees.

N = int(1e5)
model = catboost.CatBoostRegressor(iterations= N, task_type="GPU", devices='0:1', max_ctr_complexity=5, 
                                   random_seed= 0, od_type='Iter', od_wait=N, verbose=int(N/10), depth=5)

_ = model.fit(X_train, y_train, eval_set=(X_test, y_test))
# clear_output()

Learning rate set to 0.008242
0:	learn: 2.9761837	test: 4.0603519	best: 4.0603519 (0)	total: 12.2ms	remaining: 20m 15s
10000:	learn: 1.8352110	test: 3.6248574	best: 3.6123507 (564)	total: 1m 51s	remaining: 16m 47s
20000:	learn: 1.8255787	test: 3.6281617	best: 3.6123507 (564)	total: 3m 38s	remaining: 14m 34s
30000:	learn: 1.8254004	test: 3.6273392	best: 3.6123507 (564)	total: 5m 26s	remaining: 12m 40s
40000:	learn: 1.8253809	test: 3.6284896	best: 3.6123507 (564)	total: 7m 13s	remaining: 10m 49s
50000:	learn: 1.8159583	test: 3.6486311	best: 3.6123507 (564)	total: 8m 58s	remaining: 8m 58s
60000:	learn: 1.8159200	test: 3.6480488	best: 3.6123507 (564)	total: 10m 48s	remaining: 7m 12s
70000:	learn: 1.7232629	test: 3.5970065	best: 3.5964915 (68758)	total: 12m 34s	remaining: 5m 23s
80000:	learn: 1.5202133	test: 3.5867399	best: 3.5736086 (75186)	total: 14m 19s	remaining: 3m 34s
90000:	learn: 1.5199647	test: 3.5862163	best: 3.5736086 (75186)	total: 15m 47s	remaining: 1m 45s
99999:	learn: 1.5175128	test: 3.5835455	best: 3.5736086 (75186)	total: 17m 13s	remaining: 0us
bestTest = 3.57360864
bestIteration = 75186
Shrink model to first 75187 iterations.

feat_importance = pd.DataFrame(list(zip(X_train.dtypes.index,
                                        model.get_feature_importance(catboost.Pool(X_train, label=y_train)))),
                               columns=['Features','Weight']).sort_values('Weight', ascending = False)

fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(14, 4))
_ = sns.barplot(y="Features", x="Weight", palette="PuRd", edgecolor = 'RoyalBlue', hatch = '///',
                data = feat_importance, ax = ax)
_ = sns.barplot(y="Features", x="Weight", facecolor = 'None', edgecolor = 'Black', data = feat_importance, ax = ax)
_ = ax.set_xlim([0, 25])
_ = ax.set_title('Feature Importance')

model.plot_tree(tree_idx=0, pool= catboost.Pool(X_train, label=y_train))

The best result for each metric calculated on each validation dataset.

display(pd.DataFrame(model.get_best_score()))

	learn	validation
RMSE	1.517513	3.573609

R2 Score

display(pd.DataFrame({'Train Set': {'R2 Score': model.score(X_train, y_train)},
 'Validation Set': {'R2 Score': model.score(X_test, y_test)}}))

	Train Set	Validation Set
R2 Score	0.716551	0.226301

---

References

1. Kaggle Dataset: Predict Future Sales
2. CatBoost Documentation

---